Version: July 6, 2021

The brittle sound of ceramics

- can vases speak?

by Mendel Kleiner1 and Paul Åström2



The possibility of recording and playing back sound by using clay as a recording medium has been studied. This study, which concerns only the case in which the clay is fired after engraving, has been made using both a simple experimental approach and a simplified theoretical approach. The results show that it may be possible to record sound into a clay surface by purely mechanical/acoustical means at a level high enough to permit detection under ideal circumstances. The detection may be done both by hearing and measurement.


In Gunnar Ekelöf's poem "Till Posthumus" in A Night in Otocac (1961, pp. 60-61), the poet directs himself to the future and suggests that one can hear

the murmur of the human past
as we hear the slogan of our youth
on an old 78-rpm record in the distance

Ekelöf's own voice and those of many contemporaries are preserved for posterity, but are there any sounds or voices preserved from antiquity?

An American electrical engineer, Richard G. Woodbridge III, 1969, wrote an article describing the first experiments in trying to establish principles of recovering sounds from antiquity. He used a crystal gramophone cartridge an connected it to a set of headphones. "The chuck of cartridge could be fitted with 'needles' of any suitable material, length, shape, etc. In all instances the cartridge was held in the fingers and could be positioned against a revolving pot mounted on a phono turntable (adjustable speed) or stroked along a paint stroke, etc." He then mounted a wheel-made pot on the phono turntable and held the phono cartridge (fitted with a needle of a sliver of wood) against the side of the revolving pot. A low-frequency chattering sound could be heard in the earphones.

Woodbridge was more successful in recovering sounds from a canvas affixed to a wooden frame. Words were spoken and music was played while the paint strokes were applied to the surface of the canvas. When the paint had dried, a wooden needle of the crystal cartridge was gently stroked in the paint and short snatches of music could be identified in the earphones.

A pot was made for Paul Åström by the Swedish artist and potter Herman Fogelin. He spoke and sang while the pot was wheeled. When the pot had been fired, the wheel-traces were reproduced in the same way as Woodbridge had done, but only the sound of the wheel could be heard.

Professor Gunnar Fant expressed his view in 1977 that it had not been strictly proven that there were possibilities of reproducing sounds from pottery.

In our work we have tried to concentrate on the possibility of recording and replaying sound by the use of clay.


Theoretical background

Clay may be well suited as a recording medium for transverse and vertical displacement. In the recording process, the wet clay has a very low mechanical impedance, i.e. the velocity resulting from a force that is applied to the clay surface is high. This means that even small forces such as those exercised by an achoustomechanichal excitation may give displacements high enough to permit later detection. Of course the clay object in practice is usually fired or dried for later use. The consequent mechanical distortion of the recorded vibration may be high if care is not taken. The mechanical distortion is possible both in large and small scale. The whole object may be deformed in the drying or firing process and the trace itself will become rougher. This introduces noise into the process.

Replaying the modulated groove is possible with a suitable transducer, either by optical or mechanical means. The interesting question to answer, in order to determine whether it is worthwhile to listen for recorded sound in antique vessels, are what the expected noise levels are in the recording/replaying process, and what the expected acoustic recording level will be.

How, then, could a modulated groove be cut into a clay surface? When forming clay objects, both hands and instruments are used. The clay is formed into the shape required and may later, while still wet, be decorated by using a chisel-like instrument such as a bird feather. Another possibility is that the fired clay object may be decorated with paint and a brush. We have only studied the first possibility here. Vibration introduced to the feather by an acoustic field due to speech, noise or music will cause vibrations in the resultant groove. Of course modulation of the groove may result from other forces, such as mechanical vibrations in the floor on which the potter's wheel is placed, and vibrations introduced by imperfect bearings or other mechanical defects.

The theory of engraving sound in the conventional disc recording medium has been described by many authors. A clear and simplified account is given by Olson (1972). In our case there is obviously no cutter in the conventional sense. A simple cutting action may be obtained by the mechanical system shown in Figure 1.

Fig. 1. Part A of this figure shows a possible engraving action on clay, obtained through the use of a feather as a decorating or forming tool. The feather and hand may be thought of as a combination of a lever and a disc, as shown in part B.

This is a simple lever system where we assume that the motion around the hub is free from friction. This mechanical system will work both as a microphone and as a mechanical transformer. The mechanical impedance ZM (=force / velocity) of the groove is determined by the properties of the clay, as well as by those of the cutting point. In order to simplify analysis, the feather or vane is assumed to be a massless lever with a transmission ratio of l2 / l1. This gives the impedance at the attachment point of the vane as ZM (l2 / l1)2. A rough estimate of the force may be obtained by assuming that the sound pickup is done as shown in Figure 2.

Fig. 2. Idealized mechanical system using the approach shown in Figure 1b.

The net force acting on the feather or vane is determined by its shape, and its size relative to the wavelength of the sound, as well as by the pressure and incidence angle of the incoming sound wave. If the vane has dimensions larger than the wavelength of the sound, the force will be at a maximum. The maximum force F on the surface will be on the order of 2 p S [N] where p is the sound pressure and S is the area of the object. This is strictly true only if the object is immovable and has dimensions which are large compared to the wavelength of sound. This means that we will have maximum force and groove modulation at frequencies which carry the consonants of speech, and which then also carry the maximum information of speech. The air surrounding the vane will try to attenuate the vibrations of the vane, but we will assume that this effect is low compared to that of the clay, assuming for the moment that the clay has a high mechanical impedance. The resulting modulation velocity v of the groove will be

vg = 2 p S / (ZM (l2 / l1))

If the dimensions of the vane are smaller than the wavelength of the incident sound, the resultant net force may be approximated by using an expression for the force of an incident plane sound wave on a solid cylinder. According to Morse (1948), the force per unit length of a cylinder can be written as


a = cylinder radius [m]
l = wavelength of sound [m]
w = 2 p f = angular frequency [r/s]
l = wavelength [m]
c = speed of sound

Using the graph given by Morse, we may obtain the approximate value of the force per unit length. If we further assume that the cylinder length l [m] is such that l >> l we may obtain a rough estimate of the force on the vane. One has to remember that if the cylinder is movable, the resultant force will be lower.

Let us assume a feather length of 30 cm, a width of 5 cm, a maximum sound pressure level of 80 dB, a frequency of 2 kHz and a lever transformation ratio (l2 / l1) of 10. The maximum force on the vane will then be

F = 1.4 p a p [N]

per unit length of the cylinder. The resultant velocity at the attachment point between the vane and the lever will be

v = F / ((l2 / l1)2 ZM) [m/s]

The modulation velocity of the groove vg will be

vg = v (l2 / l1) [m/s]

In the case which we have assumed above, the resultant velocity will be

vg = 6.6 .10-2 / ZM [m/s]

which produces a groove modulation xg of

xg = vg / 2 p f [m]

if the mechanical impedance of the groove is high, and if the vane and lever are massless and can be moved without losses. It is now primarily interesting to study the mechanical impedance of the fired groove in order to determine the possibility of detecting sound modulated grooves in fired clay surfaces.



Experiments were done both in order to determine the mechanical impedance of the clay in the cutting action and to determine the noise levels.

The mechanical impedance of the clay was determined by measurement of the force acting on a cutting needle due to linear motion of the clay. An experimental setup like the one shown in Figure 3 was used for the purpose.

Fig. 3. Principle of measurement for measuring the mechanical impedance (force / velocity) of a clay surface.

The friction of the small model railroad car was determined as insignificant compared to the action of the clay on the cutting needles. It is likely that the mechanical impedance of the clay will depend on the speed of the clay surface relative to the cutting point, and on the geometry of the cutting point. However, for the sake of simplicity, we have assumed independence of those factors in our analysis.

In order to determine the noise levels of a typical clay groove, the following approach was used. A cylinder of clay was formed and attached to a carrying cylinder which could be rotated. An old Edison wax dictaphone unit was used for this purpose. Modulation of the groove was accomplished by using a recording cutter head for 78 rpm records.

The modulation was in the plane of the clay surface and perpendicular to the groove. The clay cylinder was then removed and fired in a kiln.

Replay of the modulated groove was done using a gramophone pickup system. A Euphonics cartridge using semiconductor benders was used, since the output voltage would then be independent of the speed used for replay. It was judged necessary to keep the replay velocity of the groove as low as possible, in order to avoid both unnecessary noise generation and destruction of the pickup needle.


Experimental results

The clay used for these experiments came from a package sold commercially for hobby purposes. The brand was Creaton, Westerwälder Tone, Iphigenie GmbH, Federal Republic of Germany. In this experiment the clay was not wetted or kneaded. The cutting needles were allowed to enter ca. 10-3 m into the clay surface, and the carriage was then dragged by means of a dynamometer to measure the force required to make two grooves. The mechanical input impedance of the clay surface on one cutter needle was measured as appoximately 2.0 Ns/m. This was for untreated clay, and it is reasonable to assume that clay which has been wetted and kneaded has a mechanical impedance which is at least one order on magnitude lower, i.e. 0.2 Ns/m. Provided that the lever and vane are rigid and massless,the behaviour of this system to the incoming wave should be roughly the same as that of the immovable cylinder.

Using the expression given above for the groove motion, we may then estimate the maximum groove modulation under the conditions given above as

xg, signal = 2.6 .10-5 [m]

When trying to form a cylinder for recordig a sound track in the clay, the clay was slightly kneaded and formed into a cylinder. The cylinder was attached to the dictaphone unit, and then turned by using a simple gramophone cutter head. Using a tone generator and amplifier, the cutter head was then used to record some modulated grooves into the clay surface. The cutter needle gave a triangular groove profile with a maximum depth of about 10-3 m. The signal used for driving the cutter was a 400 Hz sine wave from a sine-wave generator. Since it was impossible to measure the amplitude of the unfired clay groove, it was assumed thet the amplitude would be unchanged except for a slight shrinking of the order of 10%.

In firing, the cylinder was slightly deformed, primarily by shrinking, but it could be reattached to the dictaphone after some filing of its internal surface. The grooves of the ceramic cylinder were replayed and the signals fed into the signal analysis equipment. Replay was accomplished a custom-made tone arm and a Euphonics U15P gramophone pickup. This pickup was used because of its piezoresistive mechanism, which made replay speed uncritical since the output signal is proportional to the deformation of the groove. The ceramic cylinder was rotated at approximately the same speed as at recording, but this is not strictly necessary. Using this pickup it would be possible to replay the groove at a much lower speed, which would tend to lower the noise as well as the abrasion of the gramophone pickup needle.

The signals were then measured as well as listened to. Listening showed that it was indeed possible to hear the signal through the noise. Figure 4 shows a part of the retrieved wave.

Fig. 4. Example of the time history of the retrieved waveform, using a phonograph cartridge and computer-based sound analysis.

The wave is contaminated by noise, and the amount of noise varied between different grooves as well as within one groove. Occasionally the reproduction may be very good as shown by Figure 5, which shows an excerpt of the waveform shown in Figure 4.

Fig. 5. Excerpt of the waveform shown in Figure 4 at about time 900 ms.

One must take into account the fact that the clay roll was formed in a very simple and unsophisticated way, and that no special action was taken to prevent it from deforming during handling and firing. It is likely that the results would have been even better if such actions had been taken.

The results of frequency analysis of the replayed signal are shown in Figure 6.

Fig. 6. Frequency analysis using FFT of the signal excerpt shown in Figure 5 (between ca. 4-15 ms). The peak at around 350 Hz corresponds to the recorded wave. Note the relative absence of noise in the frequency range of interest for speech, i.e. at 1-2 kHz.

The peak at approximately 350 Hz is due to the signal, and the rest of the spectrum is noise. The signal-to-noise ratio is approximately 0 dB measured over the frequency range of 0.1 kHz to 4 kHz. This corresponds well with the listening results. Calibration of the Euphonics pickup showed that the amplitude of the groove in the ceramic cylinder was approximately 8.10-4 m. This means that the noise level of the groove was approximately on the order of

xg, noise = 8.10-4 [m]

as measured over the frequency range of 0.1 to 4 kHz.



If one compares the level of the noise modulation in our experimental cylinder with the level of the theoretical groove modulation possible by the sound wave as calculated above, one finds that the level recorded by acoustical means (using our optimistic approximation) may in fact be of the same order. This indicates that further experiments and a more exact analysis should be of interest.


Ekelöf, G. 1961. A Night in Otocac. Albert Bonniers förlag, Stockholm, 60-61.

Morse, P. M. 1948. Vibration and Sound. McGraw-Hill Book Co., New York, 352-353.

Olson, H. F. 1972. Modern sound Reproduction. Van Nostrand Reinhold Co., New York, 169-170.

Woodbridge, R. G. III. 1969. "Acoustic Recordings from Antiquity", IEEE Proceedings, 1465-1466.


1. Mendel Kleiner, Department of Applied Acoustics, Chalmers University of Technology, S-412 96 Göteborg, Sweden.

2. Paul Åström (1929 - 2008), Department of Ancient Culture and Civilization, Institute of Classical Studies, University of Göteborg, S-411 17 Göteborg, Sweden.


This article was originally published in:
Archaeology and Natural Science 1 (1993), Jonsered 1993, pp. 66-72.


Main page